@inproceedings{ma-etal-2025-improving,
    title = "Improving Preference Alignment of {LLM} with Inference-Free Self-Refinement",
    author = "Ma, Fukun  and
      Tian, Kaibin  and
      Xue, Jieting  and
      Wang, Xiaoyi  and
      Ma, Ye  and
      Chen, Quan  and
      Jiang, Peng  and
      Wen, Lijie",
    editor = "Christodoulopoulos, Christos  and
      Chakraborty, Tanmoy  and
      Rose, Carolyn  and
      Peng, Violet",
    booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2025",
    month = nov,
    year = "2025",
    address = "Suzhou, China",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/name-variant-enfa-fane/2025.findings-emnlp.1329/",
    doi = "10.18653/v1/2025.findings-emnlp.1329",
    pages = "24459--24473",
    ISBN = "979-8-89176-335-7",
    abstract = "Large language models (LLMs) develop the in-context learning capability through pretraining and instruction tuning, enabling task adaptation without parameter updates. Self-refinement is a manifestation of this capability, which allows LLMs to iteratively refine the output using self-generated feedback. However, empirical observations reveal Inference-Free Self-Refinement (IFSR) in preference alignment: LLMs generate preference-improved output via fixed instructions, requiring no specific feedback, even no initial responses. There are two key components of the IFSR in preference alignment. The refining instruction is a fixed instruction that constrains the output distribution from a preference-semantic perspective. During training, it facilitates joint learning of preference-related semantic representations and data distribution alignment. The pseudo reference response is constructed from paired preference data and serves as a demonstration to guide the output distribution. It mitigates off-policy distributional bias while enhancing token-level preference learning in training. Experiments across multiple datasets demonstrate that incorporating IFSR into preference alignment yields performance improvement over 10{\%}. Further ablation studies reveal additional characteristics and potential principles of IFSR."
}